For a detailed reliability test this work package aims for common data analyses using the complementary information of TAIGA and KASCADE, in particular to apply deep learning (machine learning) methods to multi-messenger analyses. Such specific analyses of the data provided by the new data centre will be performed to test the entire concept resulting in full reviewed journal publications. This will give important contributions and confidence to the project as a valuable scientific tool. In addition, having young scientists working directly with the provided hard- and software environment will be the best advertisement of the infrastructure to the community.
The KASCADE and TAIGA experiments have already shown significant achievements in using the experimental data for explaining the origin of high-energy cosmic rays and gamma rays. We propose to perform within this project a full analysis of the data provided by KCDC (TAIGA and KASCADE) to test the entire data centre and to show the proof-of-principle that a real scientific multi-messenger analysis is possible by using data from a public data centre. Until now, no one performed the physics analysis with an entire data set in the wide energy range of cosmic rays.
A scientific example could be the search for the hidden hot spots and asymmetry on the sky map by combination the data from both experiments. Since KASCADE and TAIGA are located in the same hemisphere, and observe almost the same part of the sky, the uncorrelated background could be rejected using multi-dimensional convolution neural networks (CNN). Simulating different populations and intensity of Galaxy accelerators, we obtain a large set of possible frames, which can be observed experimentally. The CNN trained on this set will allow filtering out background and classifying all possible hidden anomalies. With careful treatment of hardware response, the statistics of Tunka (10 years) can be add to the statistics of KASCADE, this way, we obtain a virtual detector with significantly improved exposure. This multivariate data analysis can restrict possible configurations of possible Galaxy accelerators of cosmic rays and PeVatrons, and test the models of the magnetic fields in the Galaxy.
Another application could be the implementation of the system for online search of transients or counterparts of electromagnetic or gravitational bursts. The model and corresponding software will be feed with the configuration of possible transients (energy, duration, location, etc.), and will 12return the possible CNN filter for the rejection of regular background. The implementation and test of this model will result a standard software library, which can be applied for the future searches.
Such analyses would derive an important combined result of various experiments such as public standard spectra of cosmic rays, which is demanded in the astroparticle physics community. Hence, the data centre is a valuable and even required part of any analysis considering data from many different experiments, i.e. multi-messenger data analyses. The data centre will be also very useful for theoreticians to interpret experimental results. A full analysis published in a reviewed journal will be performed entirely based on data provided by the public data centre. This will give important input to the consolidation of the data centre.